Muscle-Bone Ratio: A New AI-Driven Biomarker for Anti-EGFR Response in Metastatic Colorectal Cancer

Highlights

• High muscle-bone ratio (MBR), automatically calculated via deep learning, is a strong predictor of progression-free survival (PFS) and overall survival (OS) benefit from panitumumab in RAS wild-type mCRC.

• The benefit of adding panitumumab to FU/FA maintenance therapy was confined to patients in the highest MBR tertiles, while sarcopenic patients showed no significant gain.

• Automated AI-based body composition analysis offers a scalable, objective method for personalizing treatment intensification in metastatic colorectal cancer.

Background: The Challenge of Heterogeneity in RAS Wild-Type mCRC

In the landscape of metastatic colorectal cancer (mCRC), the management of RAS wild-type (WT) disease has been significantly advanced by the introduction of anti-epidermal growth factor receptor (EGFR) therapies such as panitumumab and cetuximab. However, despite molecular stratification, clinical responses remain heterogeneous. Current treatment guidelines primarily rely on tumor-intrinsic factors, such as RAS/BRAF mutation status and primary tumor location (left-sided vs. right-sided). Yet, these factors do not fully account for host-related physiological variations that influence drug metabolism, systemic inflammation, and overall treatment tolerance.

Sarcopenia, the progressive loss of skeletal muscle mass and strength, has emerged as a critical host-related factor in oncology. It is associated with increased chemotherapy toxicity, surgical complications, and poor survival outcomes. Despite its clinical relevance, sarcopenia assessment has historically been hindered by the labor-intensive nature of manual CT image segmentation. The advent of deep learning now allows for the automated, high-throughput quantification of body composition markers, such as the muscle-bone ratio (MBR). This study by Keyl et al. investigates whether these AI-derived markers can move beyond prognosticating survival to actually predicting the therapeutic benefit of anti-EGFR intensification.

Study Design: Integrating AI into Clinical Trial Analysis

The researchers utilized data from the prospective PanaMa study (AIO KRK 0212; NCT01991873), a randomized phase II trial that evaluated maintenance therapy in patients with RAS WT mCRC. After receiving induction therapy with 5-fluorouracil, folinic acid, and oxaliplatin (mFOLFOX6) plus panitumumab, patients were randomized to maintenance with FU/FA alone or FU/FA plus panitumumab.

The primary innovation of this analysis was the use of a validated deep learning model to automatically calculate the MBR from baseline CT images. MBR is defined as the ratio of the skeletal muscle area to the bone area at the level of the third lumbar vertebra (L3). This ratio provides a normalized measure of muscle mass that accounts for skeletal size. Patients were stratified into tertiles based on their MBR. The primary endpoints were progression-free survival (PFS) and overall survival (OS). To ensure the robustness of the findings, the researchers also utilized a retrospective real-world validation cohort of mCRC patients treated with cetuximab.

Key Findings: MBR as a Predictive Biomarker for Anti-EGFR Therapy

PFS and OS Outcomes in the PanaMa Cohort

Of the 248 randomized patients in the PanaMa trial, pre-maintenance CT images suitable for AI analysis were available for 189 (76.2%). The results demonstrated a striking interaction between MBR and the efficacy of panitumumab. In the group receiving FU/FA plus panitumumab, patients with a high MBR (top tertiles) experienced significantly longer PFS compared to those with low MBR (HR 0.43, 95% CI: 0.25-0.73, P=0.002). This translated into an OS benefit as well (HR 0.41, 95% CI: 0.21-0.77, P=0.006).

Crucially, in patients receiving FU/FA alone, MBR was not significantly associated with outcomes, suggesting that MBR is specifically predictive of the response to anti-EGFR therapy rather than just being a general prognostic indicator of health. When comparing the two treatment arms, panitumumab provided a significant PFS benefit only in patients with a high MBR (HR 0.42, 95% CI: 0.24-0.73, P=0.002). Patients with low MBR did not derive a statistically significant benefit from the addition of panitumumab to their maintenance regimen.

Real-World Validation

The findings were further corroborated in an independent real-world cohort of patients treated with cetuximab. In this group, high MBR was again associated with superior PFS (P=0.002) and OS (P<0.001). This validation across different anti-EGFR agents (panitumumab and cetuximab) and different clinical settings (trial vs. real-world) reinforces the potential of MBR as a reliable clinical biomarker.

Expert Commentary: Mechanistic Insights and Clinical Implications

Biological Plausibility: Why Does Muscle Mass Matter?

The association between muscle mass and anti-EGFR efficacy is likely multifactorial. Skeletal muscle is not merely a structural tissue; it is a metabolically active organ that secretes myokines and influences systemic inflammation. Sarcopenia is often a marker of a pro-inflammatory state and altered drug pharmacokinetics. Patients with low muscle mass may experience higher relative drug concentrations and increased off-target toxicities, which can lead to dose reductions or treatment interruptions, ultimately compromising efficacy.

Furthermore, the EGFR signaling pathway is involved in muscle homeostasis. It is possible that the systemic milieu associated with sarcopenia interferes with the molecular mechanisms through which anti-EGFR antibodies exert their anti-tumor effects. Understanding these host-tumor interactions is essential for the next generation of precision oncology.

Limitations and Future Directions

While these results are compelling, the study has limitations. The analysis was retrospective in nature, even though it used data from a prospective trial. Additionally, while the deep learning model is validated, the specific MBR cut-off values for different populations and ethnicities still need to be standardized. Future prospective trials should consider incorporating body composition analysis as a pre-planned stratification factor to confirm these findings.

Conclusion: Moving Toward Precision Sarcopenia Assessment

The study by Keyl et al. represents a significant step forward in integrating host-related factors into the oncology decision-making process. By demonstrating that the benefit of anti-EGFR therapy in mCRC is largely confined to patients with higher muscle mass, the researchers have provided a potential tool to avoid over-treatment in sarcopenic patients who are unlikely to benefit from intensification. As AI-driven automated CT analysis becomes more integrated into radiology workflows, markers like MBR could become as routine as assessing RAS status, leading to more personalized and effective care for patients with metastatic colorectal cancer.

Funding and Clinical Trial Information

The PanaMa study (AIO KRK 0212) was supported by Amgen and various academic grants. ClinicalTrials.gov Identifier: NCT01991873. The deep learning work was supported by institutional research funds dedicated to medical AI development.

References

1. Keyl J, et al. Deep Learning-derived Sarcopenia Marker Predicts Benefit from Anti-EGFR Therapy in Patients with RAS Wild-Type Metastatic Colorectal Cancer. Clin Cancer Res. 2026; doi: 10.1158/1078-0432.CCR-25-3080.

2. Modest DP, et al. Panitumumab Maintenance Therapy in Patients With RAS Wild-Type Metastatic Colorectal Cancer: A Randomized Phase 2 Study (PanaMa). J Clin Oncol. 2022;40(1):72-82.

3. Fearon K, et al. Definition and classification of cancer cachexia: an international consensus. Lancet Oncol. 2011;12(5):489-495.